export in real JSON

[dbrrt]

Currently using deepdiff for reporting different between objects versioned, I store the return of this function using pymongo, but some of the attributes values are not JSON compliant and I can't really store them without converting these values, for instance, when I add new dictionaries to my object, the set value added will not be serializable as is (I've to convert these as lists). It'd be nice if we could have an option to output JSON compliant object.

[seperman]

@dbrrt Good point. We will make it json serializable.

[dbrrt]

For those who have the same issue, I bypassed this, by creating a custom encoder Python Class,

class SetEncoder(json.JSONEncoder):
    def default(self, obj):
       if isinstance(obj, set):
          return list(obj)
       return json.JSONEncoder.default(self, obj)

And I'm invoking it, like this :

from utilities import SetEncoder

delta['diff'] = json.dumps(DeepDiff(oldjson,newjson,
                                                exclude_paths={"root['creation']",
                                                               "root['_id']",
                                                               "root['version']",
                                                               "root['last_version']"}), cls=SetEncoder)

[seperman]

We released the version 3 today. It has a more complex data structure now since we offer the tree view. It is much easier to json dump the text view than tree view. I have to think about how to serialize the tree view.

[dbrrt]

Thanks for the notification, I'll check-up that

[robertglen]

Ironic, I had just discovered, coded against, and deployed this to my production fleet yesterday as 2.5.3 probably moments before 3.0 was released.

Today I have run into the non-JSON serializable issues for datetime objects, int types, and bool types. I can work around the datetime objects by using flask jsonify which handles and converts datetime to a str, but I'm screwed when values are a literal type instead of an instance of the type such as int and bool.

v3.5 milestone you say? would it be sufficient to do an isinstance(value, type) and pass the value through str() if that is True? Just to make it work.

[dbrrt]

@robertglen To avoid any trouble with Datetime Objects, I only use the ISO String format representation, which is really convenient, because any time manipulation library, native or not, can parse it; and yes you can use a workaround before it'll be plainly supported by DeepDiff, by overriding a JSON encoder as I did above. At least, it works.

[robertglen]

@dbrrt I've implemented your encoder class with an added ability to handle type objects as well as datetime.datetime. All is well, thanks.

[seperman]

@robertglen Can you please post your implementation here or make a PR?
I will add that to DeepDiff and release 3.0.1 immediately. Thanks.

[seperman]

This issue is resolved: https://github.com/seperman/deepdiff#serialization